PulseAugur
中
实时 22:04:21
English(EN) The Illusion of Power Capping in LLM Decode: A Phase-Aware Energy Characterisation Across Attention Architectures

研究发现LLM解码中的功率上限无效

一篇新的研究论文对在大型语言模型(LLM)服务中对GPU进行功率上限设置的有效性提出了质疑,尤其是在自回归解码阶段。研究发现,功率上限对解码能耗影响甚微,因为GPU经常处于空闲状态,等待高带宽内存(HBM)。研究人员建议,锁定SM时钟是减少功耗的更有效方法,同时不会显著降低吞吐量。该论文还指出,与传统的注意力机制相比,Mamba2等循环架构可以在预填充和解码阶段提供节能效果。 AI

影响 提出了优化LLM服务能耗的替代方法,可能影响基础设施成本和效率。

排序理由 该集群包含一篇详细介绍LLM基础设施和能效新发现的研究论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →

研究发现LLM解码中的功率上限无效

本文如何被排名

Signal score
0 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
该集群包含一篇详细介绍LLM基础设施和能效新发现的研究论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
paper, infra
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
3 days old
Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

报道来源 [1]

  1. arXiv cs.AI TIER_1 English(EN) · Bole Ma, Ayesha Afzal, Jan Eitzinger, Gerhard Wellein ·

    LLM解码中权力上限的幻觉:跨注意力架构的感知相位能量表征

    arXiv:2605.11999v2 Announce Type: replace-cross Abstract: Power capping is a common GPU energy lever in LLM serving, and it appears to work: the cap is set, the GPU reports drawing less than it, and the energy budget is met. We show the cap does little for the phase that dominate…